After the blamed commit, IPv4 SYN packets handled
by a dual stack IPv6 socket are dropped, even if
perfectly valid.
$ nstat | grep MD5
TcpExtTCPMD5Failure             5                  0.0
For a dual stack listener, an incoming IPv4 SYN packet
would call tcp_inbound_md5_hash() with @family == AF_INET,
while tp->af_specific is pointing to tcp_sock_ipv6_specific.
Only later when an IPv4-mapped child is created, tp->af_specific
is changed to tcp_sock_ipv6_mapped_specific.
Fixes: 7bbb765b7349 ("net/tcp: Merge TCP-MD5 inbound callbacks")
Reported-by: Brian Vazquez <brianvv@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Dmitry Safonov <dima@arista.com>
Tested-by: Leonard Crestez <cdleonard@gmail.com>
Link: https://lore.kernel.org/r/20220726115743.2759832-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
 
                return SKB_DROP_REASON_TCP_MD5UNEXPECTED;
        }
 
-       /* check the signature */
-       genhash = tp->af_specific->calc_md5_hash(newhash, hash_expected,
-                                                NULL, skb);
+       /* Check the signature.
+        * To support dual stack listeners, we need to handle
+        * IPv4-mapped case.
+        */
+       if (family == AF_INET)
+               genhash = tcp_v4_md5_hash_skb(newhash,
+                                             hash_expected,
+                                             NULL, skb);
+       else
+               genhash = tp->af_specific->calc_md5_hash(newhash,
+                                                        hash_expected,
+                                                        NULL, skb);
 
        if (genhash || memcmp(hash_location, newhash, 16) != 0) {
                NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);